Croatia’s mandatory pension funds hold EUR 29.1 billion. That is about a third of the country’s annual GDP, roughly the value of every share listed on the Zagreb Stock Exchange, and it belongs to about two and half million people, most of whom will not see a cent of it for decades. The number gets quoted every month. What rarely gets asked is where it came from and where it went: how much of it was ever paid in by anyone, and what the money actually bought once it arrived. Both questions can be answered from published HANFA and Zagreb Stock Exchange tables, with no estimates and no guesswork, so this post does exactly that.

The change in a fund’s net assets in any year is the sum of three things: net contributions paid in, payouts paid out when members’ accounts are closed, and the investment result on everything already in the fund. HANFA publishes the first two in its monthly reports (tables 1.5 and 1.7), so the third follows as a residual: change in net assets less contributions plus payouts. That residual is the system’s investment result in the broadest sense, after all fees: interest, dividends, realised and unrealised gains, and FX effects. HANFA has published payouts since 2019, so the decomposition below starts at the end of 2018; for the full history since 2002 HANFA publishes cumulative totals, used further below.

Second pillar net assets bridge, 31 Dec 2018 – 31 Aug 2026 (EURbn)

Source: HANFA, InterCapital Research

Between the end of 2018 and August 2026 net assets grew from EUR 13.0 billion to EUR 29.1 billion. Members paid in EUR 9.7 billion of net contributions and EUR 2.5 billion was paid out on account closure, a net inflow of EUR 7.2 billion. The remaining EUR 8.9 billion, 55 % of the growth, is investment result. It did not arrive evenly: EUR 2.4 billion over 2019–2021, a loss of EUR 0.9 billion in 2022, the only year in which net assets fell, and EUR 7.4 billion over 2023–August 2026.

Year by year, the funds earned more than they collected in six of the eight periods. The exceptions are 2020, when the investment result was EUR 152 million, about 1 % of opening assets, and 2022, when the funds collected EUR 812 million net of payouts and still ended the year EUR 107 million smaller, because the investment result was minus EUR 919 million, about 5 % of opening assets due to bad market conditions. In 2023, 2024 and 2025 the investment result was between EUR 1.7 and 2.0 billion a year, one and a half to two times the net inflow, and in the first eight months of 2026 it was EUR 1.8 billion against EUR 0.9 billion of net inflow.

Second pillar change in net assets by component (2019 – Jan-Aug 2026, EURm)

Source: HANFA, InterCapital Research

The two flows could hardly be more different in character. The net inflow is the economy showing up in the accounts: it tracks employment and wages, and it is almost boringly steady, between 4.6 % and 5.7 % of opening assets in every single year from 2019 to 2025, pandemic included. The investment result is market risk showing up in the same accounts: over the same years it swung from minus 5 % to plus 10 % of opening assets, a range three times wider than the inflow’s, and it is the line that turned 2022 into the only year of decline and 2023–2025 into the best stretch in the system’s history.

Underneath the net inflow, both components are rising. Net contributions went from EUR 889 million in 2019 to EUR 1,751 million in 2025, roughly doubling, as employment and wages both rapidly grew. Payouts on account closure went from EUR 151 million to EUR 503 million over the same years, from 17 % of contributions to 29 %, and were 31 % of contributions in January–August 2026. None of this is yet a mature system. Membership was compulsory only for people under 40 in 2002, so the first generation for whom the second pillar was never optional, those born in 1962, reaches the statutory retirement age of 65 next year, in 2027. Everyone who has drawn a second-pillar pension so far was either an older voluntary entrant or an early retiree. The real payout phase has not started; it is about to. When it does, a system that today adds EUR 1.5 billion of fresh money to markets every year will gradually turn into one that takes money out. Payouts at under a third of contributions say that moment is still well over a decade away. The trend from 17 % to 31 % in seven years says it is coming, slowly, but coming.

Measured from the start of the system in 2002, the EUR 29.1 billion in the funds today consists of EUR 19.4 billion of net contributions, less EUR 3.5 billion already paid out, and EUR 13.3 billion of cumulative investment result. Put differently: of every ten euros in the second pillar, workers and employers paid in a little over five, and markets added the rest. That ratio is a property of the system, not of any one account. A member who entered in 2002 has had two decades of compounding; one who started work last year has had none. But at the level of the country, the second pillar has by now earned nearly as much as it has collected, and that is a fact that tends to get lost in debates about contribution rates.

Before looking at where the money went, it helps to put the inflow next to the market it is often expected to support.

Second pillar net contributions vs. ZSE equity turnover (2014-Jan-Aug 2026, EURm)

Source: HANFA, ZSE, InterCapital Research

In every year from 2014 to 2025, the net contributions paid into second-pillar funds exceeded the total annual turnover in shares on the Zagreb Stock Exchange, order book and block trades combined. The ratio was 1.6 times in 2014, around 3 times in 2021–2022, 4 times in 2023–2024, and 2.3 times in 2025, a year in which equity turnover more than doubled to EUR 765 million. Over the twelve years, EUR 12.0 billion of contributions were paid into the funds while EUR 4.8 billion worth of shares changed hands on the exchange, about a quarter of that in block trades outside the order book. The first eight months of 2026 continue the pattern at the higher level of activity: EUR 1,256 million of net contributions against EUR 585 million of equity turnover, 2.1 times, with turnover already three quarters of the full-year 2025 figure.

This is a statement about size, not about intent. Contributions are not earmarked for domestic equities, and the funds operate under concentration and diversification limits. But the arithmetic is unforgiving: a market that trades in a whole year less than half of what the funds collect in that year cannot absorb the marginal euro, however willing the buyer. The marginal euro therefore goes somewhere else. Where, exactly, is the second question.

HANFA’s table C-4 shows the investment structure of the funds by asset class, domestic and foreign, at each month-end. The simplest cut is the split between domestic and foreign assets.

Domestic vs. foreign assets of mandatory pension funds (Aug 2014 – Aug 2026, % of net assets)

Source: HANFA, InterCapital Research

When HANFA’s series began in August 2014, domestic assets were 86 % of net assets and foreign assets 13 %. All figures in this post sum the three fund categories, but the picture is in practice the picture of category B: the default, middle-risk category holds EUR 25.8 billion, 88 % of the system, while the equity-heavier category A and the bond-heavy category C, meant for members furthest from and closest to retirement respectively, hold about EUR 1.9 billion and EUR 1.5 billion between them.The two lines barely moved for six years: at the end of 2019 the split was still 85 % to 14 %. The change is concentrated in the period after 2020, and by August 2026 the split was 54 % to 45 %. The system is still predominantly domestic, but the gap has closed from more than 70 percentage points to under 10.

Mandatory pension funds investment structure at year-end (2014 – Aug 2026, % of net assets)

Source: HANFA, InterCapital Research

The year-end structure shows which lines moved. At the end of 2014 the second pillar was essentially a holder of Croatian government bonds: 72 % of net assets, with domestic equities at 10 %, foreign equities at 8 % and everything else in single digits. The government bond share held between 68 % and 73 % in every year to 2019, was still 58 % at the end of 2022, and then fell to 53 %, 44 %, 38 % and 35 % by August 2026. Foreign government bonds were below 5 % of net assets at the end of 2022 and 19 % by August 2026. The other lines moved far less: domestic equities from 10 % to 15 %, foreign equities from 8 % to 12 %, foreign funds and corporate paper combined from 4 % to 15 %.

Three things changed around 2023, and none of them was a decision taken in any fund’s investment committee. The first is regulatory. The Mandatory Pension Funds Act requires that a minimum share of a fund’s assets, 60 % for category B funds and 90 % for category C, be held in the currency in which pensions are paid. Until the end of 2022 that currency was the kuna, and the only sizeable kuna-denominated asset class was Croatian government debt; in practice the rule was a domestic bond requirement. From 1 January 2023 the currency of payment is the euro, and any euro-denominated sovereign bond satisfies the same rule, whether issued by Croatia, by another euro-area member or by a non-euro EU member borrowing in euros. The second is price. Croatia’s sovereign rating moved from BBB- in 2019 to A, and Croatian yields converged on the euro-area core to the point where, at times, the euro-denominated bonds of some non-euro EU members, Poland among them, offered more than Croatian paper with a comparable or better rating. For a euro investor that is simply a better price for similar risk, in markets that are many times deeper. The third is the domestic auction itself. Croatian banks, sitting on excess liquidity since euro adoption, have become the dominant bidders for new domestic government issues, and the state has in parallel placed part of its borrowing with households through retail bonds and bills. The funds have not been shut out of that market; they have simply not needed to be the marginal buyer in it, and the price on offer elsewhere has done the rest.

The domestic lines did not shrink in absolute terms; they stopped growing as fast as the total. Holdings of Croatian government bonds rose from EUR 6.3 billion to EUR 10.2 billion, and domestic equities and GDRs from EUR 0.9 billion to EUR 4.2 billion. The funds hold more of both than ever. What changed is where the growth went.

Change in mandatory pension funds holdings by segment (31 Dec 2014 – 31 Aug 2026, EURm)

Source: HANFA, InterCapital Research

Net assets grew by EUR 20.3 billion between the two dates. The single largest destination was foreign government bonds, up EUR 5.4 billion from a base of EUR 62 million. Domestic government bonds added EUR 3.9 billion, domestic equities EUR 3.3 billion, foreign equities EUR 2.7 billion and foreign UCITS/ETFs EUR 2.4 billion; foreign corporate paper and foreign alternative funds, both zero in 2014, added EUR 0.8 billion and EUR 0.7 billion. Foreign segments account for about 60 % of the growth, domestic segments for about 40 %.

Read together, the two government-bond lines describe the mechanism. Sovereign paper is still more than half of the portfolio, but an increasing share of it is foreign. The equity side, by contrast, has stayed predominantly domestic: domestic equities are still larger than foreign equities, and the ratio between the two has barely moved in twelve years.

A change in a holding is the sum of net purchases and the change in the value of what was already held, including currency effects on foreign positions. HANFA’s aggregate tables do not separate the two, and this post does not try to. For segments that started from zero or near zero, such as foreign government bonds, foreign corporate paper and foreign alternative funds, the increase is by definition the result of purchases. For segments with a large opening position, such as domestic equities, which rose 4.7 times, or domestic government bonds, the aggregate alone cannot say how much was bought and how much was revalued.

Answering that would require position-level data for individual funds and an assessment of their decisions, which is outside the purpose of this note. The point here is the shape of the system as a whole: what came in, what was earned, and in which segments the balance sheet grew. On that level the data are complete and published, and no estimates are needed.

None of this is news. The second pillar has been too large for its home market for years; what the numbers add is the cause and the consequence in one place. The funds collect in a year more than twice what the Zagreb exchange trades in one, so the growth went abroad, into foreign bonds and foreign equities alike. Domestic equities held their ground; it was the domestic government bond share that gave way, and the timing follows the euro, the rating path and the pricing of other sovereigns. The money to keep more of it at home clearly exists. What is missing is paper worth buying: new listings of real size and quality, enough of them to absorb even a fraction of EUR 1.5 billion of fresh contributions a year. The constraint is not on the demand side of the Zagreb Stock Exchange. We welcome every quality listing that comes to market, and hope there will be many more of them in the years ahead.

Data note. Net assets, net contributions and payouts on account closure are taken from HANFA’s monthly reports on pension funds (December issue for each year, August 2026 for the current year), tables 1.5, 1.7 and 1.8 (tables 7 and 9 in the pre-2019 format). Investment result = change in net assets − net contributions + payouts; shown for 2019 onwards, the period for which HANFA publishes payouts, and cumulatively since 2002 from the “since inception” columns of the same tables. Investment structure is HANFA table C-4, categories A, B and C summed. Values published in kuna are converted at the fixed conversion rate of 7.53450. ZSE turnover is from the exchange’s annual trading reports and turnover statistics; January–August 2026 from the exchange’s monthly trading report for August 2026 (year-to-date columns), so that both sides of the comparison cover the same period. The currency-matching requirement is Article 129 of the Mandatory Pension Funds Act. Transfers between funds when a member switches are excluded, as they net to zero for the system.